Comparative Pass Analysis between Aleix Garcia and Kirian Rodriguez in the 2023/2024 Season¶

In this report, we carry out a detailed analysis of the two central midfielders with the highest number of passes in the Spanish League during the 2023/2024 season: Aleix Garcia and Kirian Rodriguez. Using a combination of passing statistics, heat maps, pass source clusters and statistical tests, we have investigated how each player distributes the ball on the pitch, identifying key patterns and differences in their playing style.

In [ ]:
import pandas as pd
import numpy as np

from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score

from scipy.stats import gaussian_kde
from statsmodels.stats.proportion import proportions_ztest

from PIL import Image
import matplotlib.pyplot as plt
import seaborn as sns


from mplsoccer import Pitch, VerticalPitch
# disable warnings
import warnings
warnings.filterwarnings('ignore')

Data loading and preparation¶

In [ ]:
passes = pd.read_csv('passes.csv')
girona_logo = 'girona.png'
las_palmas_logo = 'udlaspalmas.png'
In [ ]:
passes.head()
Out[ ]:
player_name team_id home_team_id team_name matchday match_name match_date home_team_score away_team_score start_frame ... outcome long_ball is_a_cross free_kick corner_kick throw_in aerial switch_of_play assist second_assist
0 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 115975 ... 0 0 0 1 0 0 1 0 0 0
1 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 110075 ... 0 1 0 0 0 0 0 0 0 0
2 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 116925 ... 1 1 1 0 1 0 1 0 1 0
3 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 134850 ... 1 0 0 0 0 0 0 0 0 0
4 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 136750 ... 1 0 0 0 0 0 0 0 0 0

5 rows × 25 columns

In [ ]:
conditions = (passes['corner_kick'] == 0) & (passes['free_kick'] == 0) & (passes['throw_in'] == 0)
filtered_passes = passes[conditions].drop(columns=['corner_kick', 'free_kick', 'throw_in'])
filtered_passes.head()
Out[ ]:
player_name team_id home_team_id team_name matchday match_name match_date home_team_score away_team_score start_frame ... start_y end_x end_y outcome long_ball is_a_cross aerial switch_of_play assist second_assist
1 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 110075 ... 68.0 67.5 13.4 0 1 0 0 0 0 0
3 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 134850 ... 73.4 79.2 70.8 1 0 0 0 0 0 0
4 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 136750 ... 22.7 35.2 6.8 1 0 0 0 0 0 0
6 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 122650 ... 51.0 31.9 43.7 1 0 0 0 0 0 0
7 Aleix Garcia 384 445 Girona FC Matchday 35 ALA - GIR 2024-05-10T21:00:00 2 2 136325 ... 48.5 27.7 52.7 1 0 0 0 0 0 0

5 rows × 22 columns

Descriptive Analysis¶

In [ ]:
cols = ['start_x', 'start_y', 'end_x', 'end_y', 'outcome']
filtered_passes.groupby('player_name')[cols].describe().transpose()
Out[ ]:
player_name Aleix Garcia Kirian Rodriguez
start_x count 2675.000000 2932.000000
mean 47.607850 48.435778
std 17.569225 16.124977
min 0.800000 2.800000
25% 34.200000 35.375000
50% 46.600000 46.800000
75% 62.400000 61.700000
max 97.800000 95.200000
start_y count 2675.000000 2932.000000
mean 50.760710 56.935846
std 23.232302 23.132820
min 1.500000 1.900000
25% 32.400000 38.000000
50% 51.900000 60.900000
75% 69.150000 75.300000
max 97.700000 99.000000
end_x count 2675.000000 2932.000000
mean 51.504636 51.389666
std 21.167515 19.728360
min 2.300000 1.400000
25% 34.600000 35.900000
50% 53.100000 52.650000
75% 67.500000 66.100000
max 100.000000 100.000000
end_y count 2675.000000 2932.000000
mean 49.566916 57.581105
std 29.490355 27.540046
min 0.000000 0.000000
25% 23.750000 35.050000
50% 49.900000 62.300000
75% 75.800000 81.200000
max 100.000000 100.000000
outcome count 2675.000000 2932.000000
mean 0.906168 0.890518
std 0.291649 0.312296
min 0.000000 0.000000
25% 1.000000 1.000000
50% 1.000000 1.000000
75% 1.000000 1.000000
max 1.000000 1.000000
In [ ]:
progressive_pass_conditions = [
    (filtered_passes['start_x'] <= 50) & (filtered_passes['end_x'] <= 50) & ((filtered_passes['end_x'] - filtered_passes['start_x']) >= 30),
    (filtered_passes['start_x'] <= 50) & (filtered_passes['end_x'] > 50) & ((filtered_passes['end_x'] - filtered_passes['start_x']) >= 15),
    (filtered_passes['start_x'] > 50) & (filtered_passes['end_x'] > 50) & ((filtered_passes['end_x'] - filtered_passes['start_x']) >= 10)
]
final_third_pass_conditions = [(filtered_passes['end_x'] > 66.6)]

# add new colums setting the conditions
filtered_passes['is_progressive'] = np.select(progressive_pass_conditions, [1, 1, 1], default=0)
filtered_passes['into_final_third'] = np.select(final_third_pass_conditions, [1], default=0)
In [ ]:
# Check descriptive statistics for the labeled columns
cols = ['is_a_cross', 'long_ball', 'switch_of_play', 'assist', 'second_assist', 'aerial', 'is_progressive', 'into_final_third','outcome']

# sum of the columns
cont = filtered_passes.groupby('player_name')[cols].sum().reset_index()
cont = cont.rename(columns={'index': 'label'})
# Add total_passes
cont['total_passes'] = filtered_passes.groupby('player_name').size().values
cont['success_prop'] = cont['outcome'] / cont['total_passes']
cont
Out[ ]:
player_name is_a_cross long_ball switch_of_play assist second_assist aerial is_progressive into_final_third outcome total_passes success_prop
0 Aleix Garcia 46 281 74 27 1 262 494 710 2424 2675 0.906168
1 Kirian Rodriguez 15 137 8 37 0 162 410 706 2611 2932 0.890518
In [ ]:
columns_to_compare = ['is_a_cross', 'long_ball', 'switch_of_play', 'assist', 'second_assist', 'aerial', 'is_progressive', 'into_final_third', 'outcome']

# Filter the data for the two players
aleix_data = filtered_passes[filtered_passes['player_name'] == 'Aleix Garcia']
kirian_data = filtered_passes[filtered_passes['player_name'] == 'Kirian Rodriguez']

# z-test for each variable
for column in columns_to_compare:
    count = cont[column].values
    nobs = cont['total_passes'].values
    z_stat, p_value = proportions_ztest(count, nobs)
    print(f"Variable: {column}, Z-statistic: {z_stat:.2f}, P-value: {p_value:.4f}")

    if p_value < 0.05:
        print('The difference is statistically significant')
    else:
        print('The difference is not statistically significant')

Variable: is_a_cross, Z-statistic: 4.36, P-value: 0.0000
The difference is statistically significant
Variable: long_ball, Z-statistic: 8.30, P-value: 0.0000
The difference is statistically significant
Variable: switch_of_play, Z-statistic: 7.77, P-value: 0.0000
The difference is statistically significant
Variable: assist, Z-statistic: -0.89, P-value: 0.3738
The difference is not statistically significant
Variable: second_assist, Z-statistic: 1.05, P-value: 0.2951
The difference is not statistically significant
Variable: aerial, Z-statistic: 6.04, P-value: 0.0000
The difference is statistically significant
Variable: is_progressive, Z-statistic: 4.56, P-value: 0.0000
The difference is statistically significant
Variable: into_final_third, Z-statistic: 2.12, P-value: 0.0340
The difference is statistically significant
Variable: outcome, Z-statistic: 1.93, P-value: 0.0531
The difference is not statistically significant

Aleix Garcia:¶

He made 2,675 passes in total with a success rate of 90.6%. He stands out for the number of long passes (281) and switches of play (74). Also notable is the number of progressive passes (494) and passes into the final third (710).

Kirian Rodriguez:¶

He made 2,932 passes in total with a success rate of 89.1%. He has fewer long passes (137) and switches of play (8) compared to Aleix. Although he has fewer progressive passes (410) and passes into the final third (706), he is quite close to Aleix in the latter.

Statistical Analysis of Differences¶

Significant Differences:¶

  • Is_a_cross: Aleix has a significantly higher number of crosses than Kirian, with a Z-statistic of 4.36 (p < 0.001).

  • Long_ball: Aleix also excels at long passes, with a Z-statistic of 8.30 (p < 0.001), indicating that this is a key aspect of his game.

  • Switch_of_play: Aleix makes significantly more switches of play than Kirian (Z-statistic of 7.77, p < 0.001).

  • Aerial: The amount of aerial passes is also significantly higher for Aleix (Z-statistic of 6.04, p < 0.001).

  • Is_progressive: Aleix has more progressive passes, suggesting a greater contribution to the team's progression (Z-statistic of 4.56, p < 0.001).

  • Into_final_third: Aleix also makes more passes into the final third, but the difference is less pronounced (Z-statistic of 2.12, p = 0.034).

Non-Significant Differences:¶

  • Assist and Second_assist: There are no significant differences between both players in terms of assists and second_assists, indicating that their direct contribution to goal scoring is comparable.

  • Outcome: Although Aleix has a slightly higher success rate, this difference is not statistically significant (Z-statistic of 1.93, p = 0.0531), suggesting that both players are consistently accurate in their passing.

Key Observations¶

Playing Style: Aleix Garcia appears to be a more long-pass, switch-play, and ball-progression oriented player, suggesting a more creative or deep-field distribution role.

Defensive and Offensive Contribution: Kirian Rodriguez, although making more total passes, appears to be more focused on maintaining possession with less emphasis on long or aerial passes, which could indicate a more conservative or holding role in midfield.

Pass Clustering¶

In [ ]:
aleix_passes = filtered_passes[filtered_passes['player_name'] == 'Aleix Garcia']
kirian_passes = filtered_passes[filtered_passes['player_name'] == 'Kirian Rodriguez']
In [ ]:
# k-means clustering for location of passes
def find_optimal_clusters(player_df, max_clusters=10):
    X = player_df[['start_x', 'start_y']]
    sil = {}
    for k in range(2, max_clusters + 1):
        kmeans = KMeans(n_clusters=k, random_state=42).fit(X)
        labels = kmeans.labels_
        sil[k] = silhouette_score(X, labels, metric='euclidean')
    return sil

for player in [aleix_passes, kirian_passes]:
    player_name = player['player_name'].values[0]
    sil = find_optimal_clusters(player)
    print(player_name, sil)

Aleix Garcia {2: 0.4134279923689161, 3: 0.3701930192260029, 4: 0.3890380534712627, 5: 0.374420437919003, 6: 0.3616458239615934, 7: 0.3438956919936499, 8: 0.33245354191049403, 9: 0.34883913259055344, 10: 0.3508896063678529}
Kirian Rodriguez {2: 0.4331493228081586, 3: 0.41297951773593256, 4: 0.3895233164239991, 5: 0.3882641354458411, 6: 0.37095761593026233, 7: 0.34319824286311823, 8: 0.3513719507271675, 9: 0.3543014142696635, 10: 0.35058650504843647}
In [ ]:
# clustering for Aleix Garcia
kmeans_aleix = KMeans(n_clusters=8, random_state=42).fit(aleix_passes[['start_x', 'start_y']])
aleix_passes.loc[:, 'cluster'] = kmeans_aleix.labels_

# clustering for Kirian Rodriguez
kmeans_kirian = KMeans(n_clusters=8, random_state=42).fit(kirian_passes[['start_x', 'start_y']])
kirian_passes.loc[:, 'cluster'] = kmeans_kirian.labels_
In [ ]:
# calculate the average location of the clusters
aleix_cluster_centers = kmeans_aleix.cluster_centers_
aleix_passes.loc[:, 'cluster_center_x'] = aleix_passes['cluster'].map(lambda x: aleix_cluster_centers[x][0])
aleix_passes.loc[:, 'cluster_center_y'] = aleix_passes['cluster'].map(lambda x: aleix_cluster_centers[x][1])

kirian_cluster_centers = kmeans_kirian.cluster_centers_
kirian_passes.loc[:,'cluster_center_x'] = kirian_passes['cluster'].map(lambda x: kirian_cluster_centers[x][0])
kirian_passes.loc[:, 'cluster_center_y'] = kirian_passes['cluster'].map(lambda x: kirian_cluster_centers[x][1])

# calculate the average location of the end points of the passes for each cluster
aleix_passes.loc[:, 'end_x_avg'] = aleix_passes.groupby('cluster')['end_x'].transform('mean')
aleix_passes.loc[:, 'end_y_avg'] = aleix_passes.groupby('cluster')['end_y'].transform('mean')

kirian_passes.loc[:, 'end_x_avg'] = kirian_passes.groupby('cluster')['end_x'].transform('mean')
kirian_passes.loc[:, 'end_y_avg'] = kirian_passes.groupby('cluster')['end_y'].transform('mean')
In [ ]:
def plot_pass_density(df, logo_path):
    # Set up the pitch
    pitch = VerticalPitch(pitch_type='opta', pitch_color='#22312b', line_color='#c7d5cc')
    fig, ax = pitch.draw(figsize=(16, 11), constrained_layout=True, tight_layout=False)
    fig.set_facecolor('#22312b')

    # Calculate the point density
    x = df['start_x']
    y = df['start_y']
    xy = np.vstack([x, y])
    z = gaussian_kde(xy)(xy)

    # Plot the density
    pitch.scatter(df['start_x'], df['start_y'], s=50, c=z, cmap='viridis', ax=ax)

    # Set the title
    ax.set_title(f'Pass Density for {df.player_name.iloc[0]}', fontsize=18, color='white')

    # Add the logo
    logo = Image.open(logo_path)    
    ax_logo = fig.add_axes([0.4, 0.85, 0.3, 0.1], anchor='NE', zorder=1)
    ax_logo.imshow(logo)
    ax_logo.axis('off')  # Hide the axis

     # Add Twitter handle text
    ax.figure.text(0.35, 0.1, '@alex_llinas13', fontsize=12, color='white', ha='left', va='bottom')


    return fig, ax


def plot_passes_by_cluster(df, logo_path, use_cluster_centers=False, legend=False, avg_end_points=False):
    # Set up the pitch
    pitch = Pitch(pitch_type='opta', pitch_color='#22312b', line_color='#c7d5cc')
    fig, ax = pitch.draw(figsize=(16, 11), constrained_layout=True, tight_layout=False)
    fig.set_facecolor('#22312b')

    # Define colors for clusters
    colors = ['#ad993c', '#ba4f45', '#6a9fb5', '#e5c07b', '#98c379',
               '#ff6347', '#4682b4', '#32cd32', '#ff69b4', '#8a2be2']
    # Plot the passes for each cluster
    for i, cluster in enumerate(df['cluster'].unique()):
        cluster_data = df[df['cluster'] == cluster]
        if use_cluster_centers:
            start_x, start_y = cluster_data['cluster_center_x'], cluster_data['cluster_center_y']
        else:
            start_x, start_y = cluster_data['start_x'], cluster_data['start_y']
        pitch.arrows(start_x, start_y, 
                     cluster_data['end_x'], cluster_data['end_y'], 
                     width=2, headwidth=10, headlength=10, 
                     color=colors[i % len(colors)], ax=ax, label=f'cluster {cluster}', zorder=1)
        
        if avg_end_points:
            pitch.arrows(start_x, start_y, 
                         cluster_data['end_x_avg'], cluster_data['end_y_avg'], 
                         width=22, headwidth=10, headlength=10, 
                         color='black', ax=ax, linestyle='dotted', label=f'cluster {cluster} avg end point', zorder=2)

    # Plot the cluster centers if the flag is set
    if use_cluster_centers:
        cluster_centers = df[['cluster_center_x', 'cluster_center_y', 'cluster']].drop_duplicates()
        for i, row in cluster_centers.iterrows():
            cluster_center_x, cluster_center_y, cluster = row['cluster_center_x'], row['cluster_center_y'], row['cluster']
            num_passes = len(df[df['cluster'] == cluster])
            marker_size = 5 + num_passes * 0.05  # Adjust the multiplier as needed
            ax.plot(cluster_center_x, cluster_center_y, 'o', color='white', markersize=marker_size)
            ax.text(cluster_center_x, cluster_center_y -1, str(num_passes), color='black', fontsize=12, ha='center')

    if legend:
        # Set up the legend
        ax.legend(facecolor='#22312b', handlelength=5, edgecolor='None', fontsize=20, loc='upper left')

    # Set the title
    ax.set_title(f'Passes by {df.player_name.iloc[0]}', fontsize=30, color='white')

    # Add the logo
    logo = Image.open(logo_path)    
    ax_logo = fig.add_axes([0.4, 0.85, 0.45, 0.1], anchor='NE', zorder=1)
    ax_logo.imshow(logo)
    ax_logo.axis('off')  # Hide the axis

     # Add Twitter handle text
    ax.figure.text(0.15, 0.12, '@alex_llinas13', fontsize=12, color='white', ha='left', va='bottom')


    return fig, ax
In [ ]:
# density plots
plot_pass_density(aleix_passes, girona_logo)
plot_pass_density(kirian_passes, las_palmas_logo)
Out[ ]:
(<Figure size 1600x1100 with 2 Axes>,
 <Axes: title={'center': 'Pass Density for Kirian Rodriguez'}>)
No description has been provided for this image
No description has been provided for this image
In [ ]:
# cluster plots
plot_passes_by_cluster(aleix_passes, girona_logo,use_cluster_centers=False)
plot_passes_by_cluster(kirian_passes, las_palmas_logo, use_cluster_centers=False)
Out[ ]:
(<Figure size 1600x1100 with 2 Axes>,
 <Axes: title={'center': 'Passes by Kirian Rodriguez'}>)
No description has been provided for this image
No description has been provided for this image
In [ ]:
plot_passes_by_cluster(aleix_passes, girona_logo, use_cluster_centers=True, avg_end_points=True)
plot_passes_by_cluster(kirian_passes, las_palmas_logo, use_cluster_centers=True, avg_end_points=True)
Out[ ]:
(<Figure size 1600x1100 with 2 Axes>,
 <Axes: title={'center': 'Passes by Kirian Rodriguez'}>)
No description has been provided for this image
No description has been provided for this image

Aleix Garcia:¶

Pass Distribution and Volume:

  • Aleix has a great distribution of passes in all the lanes of the field, both in his own half and in the opponent's.

  • The clusters with the highest number of passes are in the centre of the field, which is typical for a central midfielder, as these players are usually the axis of distribution.

Average Pass Direction and Length:

  • Passes from the centre of the field to the wing and between the wings are quite frequent and seem longer than those of Kirian.

  • In the centre-right lane of the field in his own half (cluster 402), the passes are more vertical but shorter, suggesting that Aleix tends to drop back to progress the game with shorter passes in this area.

Kirian Rodriguez:¶

Pass Distribution and Volume:

  • Kirian also shows a wide distribution, but with a slight preference for participating in the clusters on the left side.

  • In the opposition half, a greater number of passes are observed in the wide lanes.

  • The placement of passes for the right central lane is significantly further back than Aleix's.

Average Pass Direction and Length:

  • Except for clusters from the right lane, passes tend to be shorter than Aleix's.

Playing Style:¶

Aleix Garcia appears to be a more dynamic central midfielder in terms of distribution, with a notable emphasis on switches of play and long passes, especially from the centre of the field to the wings. This is consistent with the statistical data above showing his higher number of long passes and switches of play.

Kirian Rodriguez, on the other hand, appears to play a more conservative role, with a focus on maintaining possession and distributing the ball from safer positions in his own half, with fewer long passes and more shorter, safe passes. He tends to operate more from the left side, both in his own field and in the opposite field, but when he moves to the right lane in attack, his passes become more deep and risky. This indicates a transition strategy, where he could be looking to exploit open spaces or connect with players in more advanced positions.

In this sense, Aleix can be key to opening up the game and changing the pace of attacks with his long passes, while Kirian could be more important for building play and keeping possession.

Clustering Analysis¶

In [ ]:
# labels for the clusters
print(aleix_cluster_centers)
aleix_passes.loc[:, 'cluster_label'] = aleix_passes['cluster'].map(
    {0: 'opp-right-center', 1: 'opp-left-wing', 2: 'own-right-center', 3: 'own-left-center',
     4: 'opp-left-center', 5: 'opp-right-wing', 6: 'own-left-wing', 7: 'own-right-wing'}
)
print(kirian_cluster_centers)
kirian_passes.loc[:, 'cluster_label'] = kirian_passes['cluster'].map(
    {0: 'opp-left-wing', 1: 'opp-right-center', 2: 'own-left-center', 3: 'own-right-center',
     4: 'opp-left-center', 5: 'own-left-wing', 6: 'opp-right-wing', 7: 'own-right-wing'}
)

[[62.21005435 37.33016304]
 [65.80625    83.52008929]
 [37.42704715 44.32655087]
 [29.19193154 63.69046455]
 [57.69311164 64.18598575]
 [67.91683502 15.56734007]
 [32.71700405 83.2562753 ]
 [33.18071895 23.58104575]]
[[64.77350835 84.26634845]
 [51.12147059 47.56823529]
 [35.43972056 66.48163673]
 [29.37754386 45.07192982]
 [64.6273399  65.65320197]
 [35.13742331 83.82177914]
 [63.25333333 23.86194444]
 [36.94813559 22.81661017]]
In [ ]:
def calculate_proportions(df, passes_df, column_name, new_column_prefix):
    # calculate the number of passes in each cluster
    pass_count = passes_df[passes_df[column_name] == 1]['cluster_label'].value_counts().reset_index()
    pass_count.columns = ['cluster_label', f'{new_column_prefix}_count']
    
    # combine the dataframes
    df = pd.merge(df, pass_count, on='cluster_label', how='outer')
    
    # fill the NaN values with 0
    df.fillna(0, inplace=True)
    df[f'{new_column_prefix}_count'] = df[f'{new_column_prefix}_count'].astype(int)
    
    # calculate the proportion of passes in each cluster
    df[f'{new_column_prefix}_prop'] = df[f'{new_column_prefix}_count'] / pass_count[f'{new_column_prefix}_count'].sum()
    
    # drop the pass count columns
    df.drop(columns=[f'{new_column_prefix}_count'], inplace=True)
    
    return df
In [ ]:
# List of variables to process
variables = [
    ('assist', 'aleix_assist', aleix_passes),
    ('assist', 'kirian_assist', kirian_passes),
    ('long_ball', 'aleix_long', aleix_passes),
    ('long_ball', 'kirian_long', kirian_passes),
    ('aerial', 'aleix_aerial', aleix_passes),
    ('aerial', 'kirian_aerial', kirian_passes),
    ('switch_of_play', 'aleix_switch_of_play', aleix_passes),
    ('switch_of_play', 'kirian_switch_of_play', kirian_passes),
    ('is_a_cross', 'aleix_cross', aleix_passes),
    ('is_a_cross', 'kirian_cross', kirian_passes),
    ('is_progressive', 'aleix_progressive', aleix_passes),
    ('is_progressive', 'kirian_progressive', kirian_passes),
    ('into_final_third', 'aleix_into_final_third', aleix_passes),
    ('into_final_third', 'kirian_into_final_third', kirian_passes)
]

# Initialize combined_df
combined_df = pd.DataFrame({'cluster_label': aleix_passes['cluster_label'].unique()})

# Process each variable
for column_name, new_column_prefix, passes_df in variables:
    combined_df = calculate_proportions(combined_df, passes_df, column_name, new_column_prefix)

# Add cluster centers to the combined_df
aleix_cluster_centers_df = aleix_passes[['cluster_center_x', 'cluster_center_y', 'cluster_label']].drop_duplicates()
kirian_cluster_centers_df = kirian_passes[['cluster_center_x', 'cluster_center_y', 'cluster_label']].drop_duplicates()

combined_df = pd.merge(combined_df, aleix_cluster_centers_df, on='cluster_label')
combined_df = pd.merge(combined_df, kirian_cluster_centers_df, on='cluster_label', suffixes=('_aleix', '_kirian'))

## CALCULATE SUCCESS RATES

# calculate the proportion of passes in each cluster
aleix_proportion = aleix_passes['cluster_label'].value_counts(normalize=True).reset_index()
aleix_proportion.columns = ['cluster_label', 'aleix_total_passes_prop']
kirian_proportion = kirian_passes['cluster_label'].value_counts(normalize=True).reset_index()
kirian_proportion.columns = ['cluster_label', 'kirian_total_passes_prop']

# calculate the number of successful passes in each cluster
aleix_success_count = aleix_passes[aleix_passes['outcome'] == 1]['cluster_label'].value_counts().reset_index()
aleix_success_count.columns = ['cluster_label', 'aleix_success_count']
kirian_success_count = kirian_passes[kirian_passes['outcome'] == 1]['cluster_label'].value_counts().reset_index()
kirian_success_count.columns = ['cluster_label', 'kirian_success_count']

# combine the two dataframes
combined_df = pd.merge(combined_df, kirian_proportion, on='cluster_label')
combined_df = pd.merge(combined_df, aleix_proportion, on='cluster_label')

combined_df = pd.merge(combined_df, aleix_success_count, on='cluster_label')
combined_df = pd.merge(combined_df, kirian_success_count, on='cluster_label')

# calculate the success rate for each cluster
combined_df['aleix_success_rate'] = combined_df['aleix_success_count'] / (combined_df['aleix_total_passes_prop'] * len(aleix_passes))
combined_df['kirian_success_rate'] = combined_df['kirian_success_count'] / (combined_df['kirian_total_passes_prop'] * len(kirian_passes))

# drop the success count columns
combined_df.drop(columns=['aleix_success_count', 'kirian_success_count'], inplace=True)
In [ ]:
# Perform the z-test for each cluster
asterisks = []
for cluster in combined_df['cluster_label'].unique():
    aleix_count = aleix_passes[aleix_passes['cluster_label'] == cluster].shape[0]
    kirian_count = kirian_passes[kirian_passes['cluster_label'] == cluster].shape[0]
    
    aleix_total = aleix_passes.shape[0]
    kirian_total = kirian_passes.shape[0]
    
    aleix_prop = aleix_count / aleix_total
    kirian_prop = kirian_count / kirian_total
    
    count = [aleix_count, kirian_count]
    nobs = [aleix_total, kirian_total]
    
    z_stat, p_value = proportions_ztest(count, nobs)
    
    print(f"Cluster: {cluster}, Aleix proportion: {aleix_prop:.2f}, Kirian proportion: {kirian_prop:.2f}, Z-statistic: {z_stat:.2f}, P-value: {p_value:.4f}")
    
    if p_value < 0.05:
        asterisks.append('*')
        print('The difference is statistically significant')
    else:
        asterisks.append('')
        print('The difference is not statistically significant')

Cluster: opp-left-center, Aleix proportion: 0.16, Kirian proportion: 0.14, Z-statistic: 2.03, P-value: 0.0420
The difference is statistically significant
Cluster: opp-left-wing, Aleix proportion: 0.08, Kirian proportion: 0.14, Z-statistic: -6.85, P-value: 0.0000
The difference is statistically significant
Cluster: opp-right-center, Aleix proportion: 0.14, Kirian proportion: 0.12, Z-statistic: 2.39, P-value: 0.0167
The difference is statistically significant
Cluster: opp-right-wing, Aleix proportion: 0.11, Kirian proportion: 0.12, Z-statistic: -1.32, P-value: 0.1860
The difference is not statistically significant
Cluster: own-left-center, Aleix proportion: 0.15, Kirian proportion: 0.17, Z-statistic: -1.87, P-value: 0.0619
The difference is not statistically significant
Cluster: own-left-wing, Aleix proportion: 0.09, Kirian proportion: 0.11, Z-statistic: -2.33, P-value: 0.0199
The difference is statistically significant
Cluster: own-right-center, Aleix proportion: 0.15, Kirian proportion: 0.10, Z-statistic: 6.05, P-value: 0.0000
The difference is statistically significant
Cluster: own-right-wing, Aleix proportion: 0.11, Kirian proportion: 0.10, Z-statistic: 1.62, P-value: 0.1045
The difference is not statistically significant
In [ ]:
# plot the proportion of passes in each cluster
fig, ax = plt.subplots(figsize=(16, 11))
width = 0.35

# bar positions
x = combined_df['cluster_label']
x_indices = range(len(x))

# bar plot
bars_aleix = ax.bar(x_indices, combined_df['aleix_total_passes_prop'], width, label='Aleix')
bars_kirina = ax.bar([i + width for i in x_indices], combined_df['kirian_total_passes_prop'], width, label='Kirian')

# labels
ax.set_xlabel('Cluster Label')
ax.set_ylabel('Proportion of passes')
ax.set_title('Proportion of passes in each cluster by player')
ax.set_xticks([i + width / 2 for i in x_indices])
ax.set_xticklabels(x)
ax.legend()

# add the asterisks
for i, bar in enumerate(bars_aleix):
    ax.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 0.01, asterisks[i], ha='center', va='bottom')

for i, bar in enumerate(bars_kirina):
    ax.text(bar.get_x() + bar.get_width() / 2, bar.get_height() + 0.01, asterisks[i], ha='center', va='bottom')

plt.show()
No description has been provided for this image

Significant Differences: Aleix Garcia has a higher proportion of passes in the opp-left-center, opp-right-center, and own-right-center zones, suggesting a greater influence in those areas, especially in the center-right of the field. Kirian Rodriguez, on the other hand, has a higher proportion in the opp-left-wing and own-left-wing zones, reflecting his tendency to operate more from the left flank, both in the opponent's and own half.

Non-Significant Differences: In the clusters where the differences are not significant (opp-right-wing, own-left-center, own-right-wing), both players appear to have a similar impact, which could indicate shared roles or team strategies that assign them similar responsibilities in these zones.

In [ ]:
combined_df.set_index('cluster_label', inplace=True)
# Rename the columns for Aleix and Kirian
aleix_cols = ['aleix_assist_prop', 'aleix_long_prop', 'aleix_aerial_prop',
              'aleix_switch_of_play_prop', 'aleix_cross_prop', 'aleix_progressive_prop',
              'aleix_into_final_third_prop', 'aleix_total_passes_prop']
kirian_cols = ['kirian_assist_prop', 'kirian_long_prop', 'kirian_aerial_prop',
               'kirian_switch_of_play_prop', 'kirian_cross_prop', 'kirian_progressive_prop',
               'kirian_into_final_third_prop', 'kirian_total_passes_prop']

aleix_rename_dict = {col: col.replace('aleix_', '').replace('_prop', '') for col in aleix_cols}
kirian_rename_dict = {col: col.replace('kirian_', '').replace('_prop', '') for col in kirian_cols}

combined_df_aleix = combined_df[aleix_cols].rename(columns=aleix_rename_dict)
combined_df_kirian = combined_df[kirian_cols].rename(columns=kirian_rename_dict)

# heatmap for Aleix
fig, ax = plt.subplots(figsize=(16, 8))
sns.heatmap(combined_df_aleix, annot=True, fmt=".2f", cmap='coolwarm', ax=ax)
ax.set_title('Aleix Pass Type Proportion Heatmap', fontsize=18)
ax.set_xlabel('Pass Type', fontsize=14)
ax.set_ylabel('Cluster', fontsize=14)
plt.xticks(rotation=0)
plt.show()

# heatmap for Kirian
fig, ax = plt.subplots(figsize=(16, 8))
sns.heatmap(combined_df_kirian, annot=True, fmt=".2f", cmap='coolwarm', ax=ax)
ax.set_title('Kirian Pass Type Proportion Heatmap', fontsize=18)
ax.set_xlabel('Pass Type', fontsize=14)
ax.set_ylabel('Cluster', fontsize=14)
plt.xticks(rotation=0)
plt.show()
No description has been provided for this image
No description has been provided for this image

Aleix Garcia:¶

Right Wing in Opponent's Half:

  • Cross: Aleix makes a high proportion of crosses (78%) from this position. This highlights his tendency to seek crosses when in this area, which is consistent with his more advanced and attacking role in this wing.
  • Long Balls and Switch of Play: Although there is a low percentage of switches of play (9%) and long passes (10%), the focus here seems to be clearly on crosses.

Right Wing in Opponent's Half:

  • Progressive and Into Final Third: Here we see a significant proportion of progressive passes (27%) and passes that advance the ball into the final third (14%), indicating that Aleix also uses this position to advance play.

Left Wing in Own Half:

  • Long Balls: There is a relevant percentage of long passes (18% and 8%) from these areas, which suggests that from his own half, Aleix Garcia tends to look for safer options or switches of play.

Central Lanes:

  • Distribution: From these positions, his distribution of pass types is more balanced, with some leaning towards long, progressive passes, but less emphasis on crosses or switches of play.

Kirian Rodriguez:¶

Left Lane in Opposition:

  • Assist and Progressive: Kirian has a high proportion of assists (41%) and progressive passes (25%) from these areas. This suggests that he is more likely to play creatively and look for scoring opportunities from the left wing in the opposition half.

  • Cross: A decent percentage of his passes from the left wing in the opposition half are also crosses (33%), although less than Aleix on the right wing.

Right Lane in Opposition

  • Cross and Switch of Play: Similar to Aleix, Kirian also makes a high proportion of crosses (60%) from this area, along with a considerable number of switches of play (23%).

Central Lane in Own Half:

  • Long Balls: From the centre of the pitch in his own half, Kirian makes a significant amount of long passes (15% and 6%), but these are less prevalent than Aleix, suggesting a more conservative distribution.

Roles and Areas of Influence:¶

Aleix Garcia: His role on the pitch seems to be very focused on being a versatile distributor, capable of making a large number of crosses from the right wing in the opposition half. He also excels at progressive passes from central areas.

Kirian Rodriguez: He tends to be more creative and dangerous from the left wing, with a high proportion of assists and progressive passes. On the right wing in the opposition half, his game becomes more similar to Aleix's, with a high frequency of crosses and switches of play.

Overall Comparison:

Aleix seems to be more balanced in his distribution of pass types, while Kirian has specific areas where certain types of passes clearly dominate, such as on the left wing in the opposition half for assists and on the right wing for crosses.

Conclusions¶

This comparative analysis reveals that, although Aleix Garcia and Kirian Rodriguez share the characteristic of being influential midfielders with a high amount of passing, their playing styles and impact zones on the pitch are different:

Aleix Garcia is a more versatile and balanced player in terms of distribution, with a preference for deep and more direct distribution of his passes.

Kirian Rodriguez is distinguished by his influence on the left flank, where he makes most of his assists and progressive passes. In addition, his tendency to make long passes from the right in his own half suggests a strategic focus on quick transitions.